Improving lapse predictions (sigma 6/2017: LIFE IN-FORCE MANAGEMENT)
Data analytics – descriptive and predictive – can be used to improve retention of inforce
business in several ways. Statistical models can be applied to investigate and
better understand who lapses and why, using data systematically collected from
different sources beyond traditional policyholder information. Based on the
understanding of the underlying drivers, predictive models can then provide
forecasts of consumer propensity to lapse in response to changes in different
variables (eg, price changes). This can be used to appropriately allocate resources
to reduce lapses.
To date, however, propensity-to-lapse models have been largely unsuccessful in
predicting lapses. New data sources and machine learning algorithms are likely
to improve their predictive power, but methods such as neural networks are like
black boxes, designed to train and predict outcome without shedding light on the
underlying decision path to lapse. A better understanding of why people lapse is
needed and to this end, traditional statistical methods remain important analytical
tools.
The life industry has been slow to implement advanced data mining and predictive
modelling techniques. According to a survey, capturing reliable data and
implementing the right solutions to analyse and interpret the data remain key
challenges for insurers.48 Insurers currently use data analytics mainly for marketing
and client segmentation purposes. However, using it for additional purposes,
including predicting lapse rates, is likely to increase as its utility becomes more
apparent.